VirtualBox

source: vbox/trunk/src/VBox/Devices/Network/slirp/slirp.c@ 19318

Last change on this file since 19318 was 19313, checked in by vboxsync, 16 years ago

NAT: slirp servicing several guests
  • Property svn:eol-style set to native
File size: 55.7 KB
Line 
1#include "slirp.h"
2#ifdef RT_OS_OS2
3# include <paths.h>
4#endif
5#ifdef VBOX_WITH_SLIRP_DNS_PROXY
6#include "dnsproxy/dnsproxy.h"
7#endif
8
9#include <VBox/err.h>
10#include <VBox/pdmdrv.h>
11#include <iprt/assert.h>
12#ifndef RT_OS_WINDOWS
13# include <sys/ioctl.h>
14# include <poll.h>
15#else
16# include <Winnls.h>
17# define _WINSOCK2API_
18# include <IPHlpApi.h>
19#endif
20
21#if !defined(RT_OS_WINDOWS)
22
23# define DO_ENGAGE_EVENT1(so, fdset, label) \
24 do { \
25 if( so->so_poll_index != -1 \
26 && so->s == polls[so->so_poll_index].fd) { \
27 polls[so->so_poll_index].events |= N_(fdset ## _poll); \
28 break; /* out of this loop */ \
29 } \
30 AssertRelease(poll_index < (nfds)); \
31 AssertRelease(poll_index >= 0 && poll_index < (nfds)); \
32 polls[poll_index].fd = (so)->s; \
33 (so)->so_poll_index = poll_index; \
34 polls[poll_index].events = N_(fdset ## _poll); \
35 polls[poll_index].revents = 0; \
36 poll_index++; \
37 } while(0)
38
39
40# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
41 do { \
42 if( so->so_poll_index != -1 \
43 && so->s == polls[so->so_poll_index].fd) { \
44 polls[so->so_poll_index].events |= \
45 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
46 break; /* out of this loop */ \
47 } \
48 AssertRelease(poll_index < (nfds)); \
49 polls[poll_index].fd = (so)->s; \
50 (so)->so_poll_index = poll_index; \
51 polls[poll_index].events = \
52 N_(fdset1 ## _poll) | N_(fdset1 ## _poll); \
53 poll_index++; \
54 } while(0)
55
56# define DO_POLL_EVENTS(rc, error, so, events, label) do {} while (0)
57
58# define DO_CHECK_FD_SET(so, events, fdset) ( ((so)->so_poll_index != -1) \
59 && ((so)->so_poll_index <= ndfs) \
60 && ((so)->s == polls[so->so_poll_index].fd) \
61 && (polls[(so)->so_poll_index].revents & N_(fdset ## _poll)))
62# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset) /*specific for Unix API */
63# define DO_WIN_CHECK_FD_SET(so, events, fdset ) 0 /* specific for Windows Winsock API */
64
65# ifndef RT_OS_WINDOWS
66
67# ifndef RT_OS_LINUX
68# define readfds_poll (POLLRDNORM)
69# define writefds_poll (POLLWRNORM)
70# define xfds_poll (POLLRDBAND|POLLWRBAND|POLLPRI)
71# else
72# define readfds_poll (POLLIN)
73# define writefds_poll (POLLOUT)
74# define xfds_poll (POLLPRI)
75# endif
76# define rderr_poll (POLLERR)
77# define rdhup_poll (POLLHUP)
78# define nval_poll (POLLNVAL)
79
80# define ICMP_ENGAGE_EVENT(so, fdset) \
81 do { \
82 if (pData->icmp_socket.s != -1) \
83 DO_ENGAGE_EVENT1((so), fdset, ICMP); \
84 } while (0)
85# else /* !RT_OS_WINDOWS */
86# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
87# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
88#endif /* RT_OS_WINDOWS */
89
90#else /* defined(RT_OS_WINDOWS) */
91
92/*
93 * On Windows, we will be notified by IcmpSendEcho2() when the response arrives.
94 * So no call to WSAEventSelect necessary.
95 */
96# define ICMP_ENGAGE_EVENT(so, fdset) do {} while(0)
97
98# define DO_ENGAGE_EVENT1(so, fdset1, label) \
99 do { \
100 rc = WSAEventSelect((so)->s, VBOX_SOCKET_EVENT, FD_ALL_EVENTS); \
101 if (rc == SOCKET_ERROR) \
102 { \
103 /* This should not happen */ \
104 error = WSAGetLastError(); \
105 LogRel(("WSAEventSelect (" #label ") error %d (so=%x, socket=%s, event=%x)\n", \
106 error, (so), (so)->s, VBOX_SOCKET_EVENT)); \
107 } \
108 } while(0); \
109 CONTINUE(label)
110
111# define DO_ENGAGE_EVENT2(so, fdset1, fdset2, label) \
112 DO_ENGAGE_EVENT1((so), (fdset1), label)
113
114# define DO_POLL_EVENTS(rc, error, so, events, label) \
115 (rc) = WSAEnumNetworkEvents((so)->s, VBOX_SOCKET_EVENT, (events)); \
116 if ((rc) == SOCKET_ERROR) \
117 { \
118 (error) = WSAGetLastError(); \
119 LogRel(("WSAEnumNetworkEvents " #label " error %d\n", (error))); \
120 CONTINUE(label); \
121 }
122
123# define acceptds_win FD_ACCEPT
124# define acceptds_win_bit FD_ACCEPT_BIT
125
126# define readfds_win FD_READ
127# define readfds_win_bit FD_READ_BIT
128
129# define writefds_win FD_WRITE
130# define writefds_win_bit FD_WRITE_BIT
131
132# define xfds_win FD_OOB
133# define xfds_win_bit FD_OOB_BIT
134
135# define DO_CHECK_FD_SET(so, events, fdset) \
136 (((events).lNetworkEvents & fdset ## _win) && ((events).iErrorCode[fdset ## _win_bit] == 0))
137
138# define DO_WIN_CHECK_FD_SET(so, events, fdset ) DO_CHECK_FD_SET((so), (events), fdset)
139# define DO_UNIX_CHECK_FD_SET(so, events, fdset ) 1 /*specific for Unix API */
140
141#endif /* defined(RT_OS_WINDOWS) */
142
143#define TCP_ENGAGE_EVENT1(so, fdset) \
144 DO_ENGAGE_EVENT1((so), fdset, tcp)
145
146#define TCP_ENGAGE_EVENT2(so, fdset1, fdset2) \
147 DO_ENGAGE_EVENT2((so), fdset1, fdset2, tcp)
148
149#define UDP_ENGAGE_EVENT(so, fdset) \
150 DO_ENGAGE_EVENT1((so), fdset, udp)
151
152#define POLL_TCP_EVENTS(rc, error, so, events) \
153 DO_POLL_EVENTS((rc), (error), (so), (events), tcp)
154
155#define POLL_UDP_EVENTS(rc, error, so, events) \
156 DO_POLL_EVENTS((rc), (error), (so), (events), udp)
157
158#define CHECK_FD_SET(so, events, set) \
159 (DO_CHECK_FD_SET((so), (events), set))
160
161#define WIN_CHECK_FD_SET(so, events, set) \
162 (DO_WIN_CHECK_FD_SET((so), (events), set))
163#define UNIX_CHECK_FD_SET(so, events, set) \
164 (DO_UNIX_CHECK_FD_SET(so, events, set))
165
166/*
167 * Loging macros
168 */
169#if VBOX_WITH_DEBUG_NAT_SOCKETS
170# if defined(RT_OS_WINDOWS)
171# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
172 do { \
173 LogRel((" " #proto " %R[natsock] %R[natwinnetevents]\n", (so), (winevent))); \
174 } while (0)
175# else /* RT_OS_WINDOWS */
176# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) \
177 do { \
178 LogRel((" " #proto " %R[natsock] %s %s %s er: %s, %s, %s\n", (so), \
179 CHECK_FD_SET(so, ign ,r_fdset) ? "READ":"", \
180 CHECK_FD_SET(so, ign, w_fdset) ? "WRITE":"", \
181 CHECK_FD_SET(so, ign, x_fdset) ? "OOB":"", \
182 CHECK_FD_SET(so, ign, rderr) ? "RDERR":"", \
183 CHECK_FD_SET(so, ign, rdhup) ? "RDHUP":"", \
184 CHECK_FD_SET(so, ign, nval) ? "RDNVAL":"")); \
185 } while (0)
186# endif /* !RT_OS_WINDOWS */
187#else /* VBOX_WITH_DEBUG_NAT_SOCKETS */
188# define DO_LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) do {} while (0)
189#endif /* !VBOX_WITH_DEBUG_NAT_SOCKETS */
190
191#define LOG_NAT_SOCK(so, proto, winevent, r_fdset, w_fdset, x_fdset) DO_LOG_NAT_SOCK((so), proto, (winevent), r_fdset, w_fdset, x_fdset)
192
193static const uint8_t special_ethaddr[6] =
194{
195 0x52, 0x54, 0x00, 0x12, 0x35, 0x00
196};
197
198static const uint8_t broadcast_ethaddr[6] =
199{
200 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
201};
202
203static const uint8_t zerro_ethaddr[6] =
204{
205 0x0, 0x0, 0x0, 0x0, 0x0, 0x0
206};
207
208#ifdef RT_OS_WINDOWS
209# ifndef VBOX_WITH_MULTI_DNS
210static int get_dns_addr_domain(PNATState pData, bool fVerbose,
211 struct in_addr *pdns_addr,
212 const char **ppszDomain)
213{
214 int rc = 0;
215 FIXED_INFO *FixedInfo = NULL;
216 ULONG BufLen;
217 DWORD ret;
218 IP_ADDR_STRING *pIPAddr;
219 struct in_addr tmp_addr;
220
221 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
222 BufLen = sizeof(FIXED_INFO);
223
224 /** @todo: this API returns all DNS servers, no matter whether the
225 * corresponding network adapter is disabled or not. Maybe replace
226 * this by GetAdapterAddresses(), which is XP/Vista only though. */
227 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen))
228 {
229 if (FixedInfo)
230 {
231 GlobalFree(FixedInfo);
232 FixedInfo = NULL;
233 }
234 FixedInfo = GlobalAlloc(GPTR, BufLen);
235 }
236
237 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS)
238 {
239 Log(("GetNetworkParams failed. ret = %08x\n", (u_int)ret ));
240 if (FixedInfo)
241 {
242 GlobalFree(FixedInfo);
243 FixedInfo = NULL;
244 }
245 rc = -1;
246 goto get_dns_prefix;
247 }
248
249 pIPAddr = &(FixedInfo->DnsServerList);
250 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
251 Log(("nat: DNS Servers:\n"));
252 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
253 LogRel(("NAT: DNS address: %s\n", pIPAddr->IpAddress.String));
254 *pdns_addr = tmp_addr;
255
256 pIPAddr = FixedInfo -> DnsServerList.Next;
257 while (pIPAddr)
258 {
259 if (fVerbose)
260 LogRel(("NAT: ignored DNS address: %s\n", pIPAddr ->IpAddress.String));
261 pIPAddr = pIPAddr ->Next;
262 }
263 if (FixedInfo)
264 {
265 GlobalFree(FixedInfo);
266 FixedInfo = NULL;
267 }
268
269get_dns_prefix:
270 if (ppszDomain)
271 {
272 OSVERSIONINFO ver;
273 char szDnsDomain[256];
274 DWORD dwSize = sizeof(szDnsDomain);
275
276 *ppszDomain = NULL;
277 GetVersionEx(&ver);
278 if (ver.dwMajorVersion >= 5)
279 {
280 /* GetComputerNameEx exists in Windows versions starting with 2000. */
281 if (GetComputerNameEx(ComputerNameDnsDomain, szDnsDomain, &dwSize))
282 {
283 if (szDnsDomain[0])
284 {
285 /* Just non-empty strings are valid. */
286 *ppszDomain = RTStrDup(szDnsDomain);
287 if (pData->fPassDomain)
288 {
289 if (fVerbose)
290 LogRel(("NAT: passing domain name %s\n", szDnsDomain));
291 }
292 else
293 Log(("nat: ignoring domain %s\n", szDnsDomain));
294 }
295 }
296 else
297 Log(("nat: GetComputerNameEx failed (%d)\n", GetLastError()));
298 }
299 }
300 return rc;
301}
302# else /* !VBOX_WITH_MULTI_DNS */
303static int get_dns_addr_domain(PNATState pData, bool fVerbose,
304 struct in_addr *pdns_addr,
305 const char **ppszDomain)
306{
307 /* Get amount of memory required for operation */
308 ULONG flags = GAA_FLAG_INCLUDE_PREFIX; /*GAA_FLAG_INCLUDE_ALL_INTERFACES;*/ /* all interfaces registered in NDIS */
309 PIP_ADAPTER_ADDRESSES addresses = NULL;
310 PIP_ADAPTER_ADDRESSES addr = NULL;
311 PIP_ADAPTER_DNS_SERVER_ADDRESS dns = NULL;
312 ULONG size = 0;
313 int wlen = 0;
314 char *suffix;
315 struct dns_entry *da = NULL;
316 struct dns_domain_entry *dd = NULL;
317 ULONG ret = ERROR_SUCCESS;
318
319 /* @todo add SKIPing flags to get only required information */
320
321 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
322 if (ret != ERROR_BUFFER_OVERFLOW)
323 {
324 LogRel(("NAT: error %lu occured on capacity detection operation\n", ret));
325 return -1;
326 }
327
328 if (size == 0)
329 {
330 LogRel(("NAT: Win socket API returns non capacity\n"));
331 return -1;
332 }
333
334 addresses = RTMemAllocZ(size);
335 if (addresses == NULL)
336 {
337 LogRel(("NAT: No memory available \n"));
338 return -1;
339 }
340
341 ret = pData->pfGetAdaptersAddresses(AF_INET, 0, NULL /* reserved */, addresses, &size);
342 if (ret != ERROR_SUCCESS)
343 {
344 LogRel(("NAT: error %lu occured on fetching adapters info\n", ret));
345 RTMemFree(addresses);
346 return -1;
347 }
348 addr = addresses;
349 while(addr != NULL)
350 {
351 int found;
352 if (addr->OperStatus != IfOperStatusUp)
353 goto next;
354 dns = addr->FirstDnsServerAddress;
355 while (dns != NULL)
356 {
357 struct sockaddr *saddr = dns->Address.lpSockaddr;
358 if (saddr->sa_family != AF_INET)
359 goto next_dns;
360 /* add dns server to list */
361 da = RTMemAllocZ(sizeof(struct dns_entry));
362 if (da == NULL)
363 {
364 LogRel(("NAT: Can't allocate buffer for DNS entry\n"));
365 RTMemFree(addresses);
366 return VERR_NO_MEMORY;
367 }
368 LogRel(("NAT: adding %R[IP4] to DNS server list\n", &((struct sockaddr_in *)saddr)->sin_addr));
369 if ((((struct sockaddr_in *)saddr)->sin_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
370 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
371 }
372 else
373 {
374 da->de_addr.s_addr = ((struct sockaddr_in *)saddr)->sin_addr.s_addr;
375 }
376 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
377
378 if (addr->DnsSuffix == NULL)
379 goto next_dns;
380
381 /*uniq*/
382 RTUtf16ToUtf8(addr->DnsSuffix, &suffix);
383
384 if (!suffix || strlen(suffix) == 0) {
385 RTStrFree(suffix);
386 goto next_dns;
387 }
388
389 found = 0;
390 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
391 {
392 if ( dd->dd_pszDomain != NULL
393 && strcmp(dd->dd_pszDomain, suffix) == 0)
394 {
395 found = 1;
396 RTStrFree(suffix);
397 break;
398 }
399 }
400 if (found == 0)
401 {
402 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
403 if (dd == NULL)
404 {
405 LogRel(("NAT: not enough memory\n"));
406 RTStrFree(suffix);
407 RTMemFree(addresses);
408 return VERR_NO_MEMORY;
409 }
410 dd->dd_pszDomain = suffix;
411 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
412 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
413 }
414 next_dns:
415 dns = dns->Next;
416 }
417 next:
418 addr = addr->Next;
419 }
420 RTMemFree(addresses);
421 return 0;
422}
423# endif /* VBOX_WITH_MULTI_DNS */
424
425#else /* !RT_OS_WINDOWS */
426
427static int get_dns_addr_domain(PNATState pData, bool fVerbose,
428 struct in_addr *pdns_addr,
429 const char **ppszDomain)
430{
431 char buff[512];
432 char buff2[256];
433 FILE *f;
434 int found = 0;
435 struct in_addr tmp_addr;
436
437#ifdef RT_OS_OS2
438 /* Try various locations. */
439 char *etc = getenv("ETC");
440 f = NULL;
441 if (etc)
442 {
443 snprintf(buff, sizeof(buff), "%s/RESOLV2", etc);
444 f = fopen(buff, "rt");
445 }
446 if (!f)
447 {
448 snprintf(buff, sizeof(buff), "%s/RESOLV2", _PATH_ETC);
449 f = fopen(buff, "rt");
450 }
451 if (!f)
452 {
453 snprintf(buff, sizeof(buff), "%s/resolv.conf", _PATH_ETC);
454 f = fopen(buff, "rt");
455 }
456#else
457 f = fopen("/etc/resolv.conf", "r");
458#endif
459 if (!f)
460 return -1;
461
462 if (ppszDomain)
463 *ppszDomain = NULL;
464 Log(("nat: DNS Servers:\n"));
465 while (fgets(buff, 512, f) != NULL)
466 {
467#ifdef VBOX_WITH_MULTI_DNS
468 struct dns_entry *da = NULL;
469#endif
470 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1)
471 {
472 if (!inet_aton(buff2, &tmp_addr))
473 continue;
474#ifndef VBOX_WITH_MULTI_DNS
475 /* If it's the first one, set it to dns_addr */
476 if (!found)
477 {
478 if (fVerbose || pdns_addr->s_addr != tmp_addr.s_addr)
479 LogRel(("NAT: DNS address: %s\n", buff2));
480 *pdns_addr = tmp_addr;
481 }
482 else
483 {
484 if (fVerbose)
485 LogRel(("NAT: ignored DNS address: %s\n", buff2));
486 }
487#else
488 /*localhost mask */
489 da = RTMemAllocZ(sizeof (struct dns_entry));
490 if (da == NULL)
491 {
492 LogRel(("can't alloc memory for DNS entry\n"));
493 return -1;
494 }
495 /*check */
496 da->de_addr.s_addr = tmp_addr.s_addr;
497 if ((da->de_addr.s_addr & htonl(IN_CLASSA_NET)) == ntohl(INADDR_LOOPBACK & IN_CLASSA_NET)) {
498 da->de_addr.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_ALIAS);
499 }
500 LIST_INSERT_HEAD(&pData->dns_list_head, da, de_list);
501#endif
502 found++;
503 }
504#ifndef VBOX_WITH_MULTI_DNS
505 if ( ppszDomain
506 && (!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
507 {
508 /* Domain name/search list present. Pick first entry */
509 if (*ppszDomain == NULL)
510 {
511 char *tok;
512 char *saveptr;
513 tok = strtok_r(&buff[6], " \t\n", &saveptr);
514 if (tok)
515 {
516 *ppszDomain = RTStrDup(tok);
517 if (pData->fPassDomain)
518 {
519 if (fVerbose)
520 LogRel(("NAT: passing domain name %s\n", tok));
521 }
522 else
523 Log(("nat: ignoring domain %s\n", tok));
524 }
525 }
526 }
527#else
528 if ((!strncmp(buff, "domain", 6) || !strncmp(buff, "search", 6)))
529 {
530 char *tok;
531 char *saveptr;
532 struct dns_domain_entry *dd = NULL;
533 int found = 0;
534 tok = strtok_r(&buff[6], " \t\n", &saveptr);
535 LIST_FOREACH(dd, &pData->dns_domain_list_head, dd_list)
536 {
537 if( tok != NULL
538 && strcmp(tok, dd->dd_pszDomain) == 0)
539 {
540 found = 1;
541 break;
542 }
543 }
544 if (tok != NULL && found == 0) {
545 dd = RTMemAllocZ(sizeof(struct dns_domain_entry));
546 if (dd == NULL)
547 {
548 LogRel(("NAT: not enought memory to add domain list\n"));
549 return VERR_NO_MEMORY;
550 }
551 dd->dd_pszDomain = RTStrDup(tok);
552 LogRel(("NAT: adding domain name %s to search list\n", dd->dd_pszDomain));
553 LIST_INSERT_HEAD(&pData->dns_domain_list_head, dd, dd_list);
554 }
555 }
556#endif
557 }
558 fclose(f);
559 if (!found)
560 return -1;
561 return 0;
562}
563
564#endif
565#ifdef VBOX_WITH_MULTI_DNS
566static int slirp_init_dns_list(PNATState pData)
567{
568 LIST_INIT(&pData->dns_list_head);
569 LIST_INIT(&pData->dns_domain_list_head);
570 return get_dns_addr_domain(pData, true, NULL, NULL);
571}
572
573static void slirp_release_dns_list(PNATState pData)
574{
575 struct dns_entry *de = NULL;
576 struct dns_domain_entry *dd = NULL;
577 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
578 dd = LIST_FIRST(&pData->dns_domain_list_head);
579 LIST_REMOVE(dd, dd_list);
580 if (dd->dd_pszDomain != NULL)
581 RTStrFree(dd->dd_pszDomain);
582 RTMemFree(dd);
583 }
584 while(!LIST_EMPTY(&pData->dns_domain_list_head)) {
585 dd = LIST_FIRST(&pData->dns_domain_list_head);
586 LIST_REMOVE(dd, dd_list);
587 if (dd->dd_pszDomain != NULL)
588 RTStrFree(dd->dd_pszDomain);
589 RTMemFree(dd);
590 }
591}
592#endif
593
594int get_dns_addr(PNATState pData, struct in_addr *pdns_addr)
595{
596 return get_dns_addr_domain(pData, false, pdns_addr, NULL);
597}
598
599#ifndef VBOX_WITH_NAT_SERVICE
600int slirp_init(PNATState *ppData, const char *pszNetAddr, uint32_t u32Netmask,
601 bool fPassDomain, void *pvUser)
602#else
603int slirp_init(PNATState *ppData, uint32_t u32NetAddr, uint32_t u32Netmask,
604 bool fPassDomain, void *pvUser)
605#endif
606{
607 int fNATfailed = 0;
608 int rc;
609 PNATState pData = RTMemAllocZ(sizeof(NATState));
610 *ppData = pData;
611 if (!pData)
612 return VERR_NO_MEMORY;
613 if (u32Netmask & 0x1f)
614 /* CTL is x.x.x.15, bootp passes up to 16 IPs (15..31) */
615 return VERR_INVALID_PARAMETER;
616 pData->fPassDomain = fPassDomain;
617 pData->pvUser = pvUser;
618 pData->netmask = u32Netmask;
619
620#ifdef RT_OS_WINDOWS
621 {
622 WSADATA Data;
623 WSAStartup(MAKEWORD(2, 0), &Data);
624 }
625 pData->phEvents[VBOX_SOCKET_EVENT_INDEX] = CreateEvent(NULL, FALSE, FALSE, NULL);
626#endif
627#ifdef VBOX_WITH_SLIRP_MT
628 QSOCKET_LOCK_CREATE(tcb);
629 QSOCKET_LOCK_CREATE(udb);
630 rc = RTReqCreateQueue(&pData->pReqQueue);
631 AssertReleaseRC(rc);
632#endif
633
634 link_up = 1;
635
636 debug_init();
637 if_init(pData);
638 ip_init(pData);
639 icmp_init(pData);
640
641 /* Initialise mbufs *after* setting the MTU */
642 m_init(pData);
643
644#ifndef VBOX_WITH_NAT_SERVICE
645 inet_aton(pszNetAddr, &special_addr);
646#else
647 special_addr.s_addr = u32NetAddr;
648#endif
649 alias_addr.s_addr = special_addr.s_addr | htonl(CTL_ALIAS);
650 /* @todo: add ability to configure this staff */
651
652 /* set default addresses */
653 inet_aton("127.0.0.1", &loopback_addr);
654#ifndef VBOX_WITH_MULTI_DNS
655 inet_aton("127.0.0.1", &dns_addr);
656
657 if (get_dns_addr_domain(pData, true, &dns_addr, &pData->pszDomain) < 0)
658#else
659 if (slirp_init_dns_list(pData) < 0)
660#endif
661 fNATfailed = 1;
662#ifdef VBOX_WITH_SLIRP_DNS_PROXY
663 dnsproxy_init(pData);
664#endif
665
666 getouraddr(pData);
667 return fNATfailed ? VINF_NAT_DNS : VINF_SUCCESS;
668}
669
670/**
671 * Statistics counters.
672 */
673void slirp_register_timers(PNATState pData, PPDMDRVINS pDrvIns)
674{
675#ifdef VBOX_WITH_STATISTICS
676 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFill, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
677 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fills", "/Drivers/NAT%d/Fill", pDrvIns->iInstance);
678 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatPoll, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
679 STAMUNIT_TICKS_PER_CALL, "Profiling slirp polls", "/Drivers/NAT%d/Poll", pDrvIns->iInstance);
680 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatFastTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
681 STAMUNIT_TICKS_PER_CALL, "Profiling slirp fast timer", "/Drivers/NAT%d/TimerFast", pDrvIns->iInstance);
682 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatSlowTimer, STAMTYPE_PROFILE, STAMVISIBILITY_ALWAYS,
683 STAMUNIT_TICKS_PER_CALL, "Profiling slirp slow timer", "/Drivers/NAT%d/TimerSlow", pDrvIns->iInstance);
684 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
685 STAMUNIT_COUNT, "TCP sockets", "/Drivers/NAT%d/SockTCP", pDrvIns->iInstance);
686 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatTCPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
687 STAMUNIT_COUNT, "TCP sockets active", "/Drivers/NAT%d/SockTCPHot", pDrvIns->iInstance);
688 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDP, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
689 STAMUNIT_COUNT, "UDP sockets", "/Drivers/NAT%d/SockUDP", pDrvIns->iInstance);
690 PDMDrvHlpSTAMRegisterF(pDrvIns, &pData->StatUDPHot, STAMTYPE_COUNTER, STAMVISIBILITY_ALWAYS,
691 STAMUNIT_COUNT, "UDP sockets active", "/Drivers/NAT%d/SockUDPHot", pDrvIns->iInstance);
692#endif /* VBOX_WITH_STATISTICS */
693}
694
695/**
696 * Marks the link as up, making it possible to establish new connections.
697 */
698void slirp_link_up(PNATState pData)
699{
700 link_up = 1;
701}
702
703/**
704 * Marks the link as down and cleans up the current connections.
705 */
706void slirp_link_down(PNATState pData)
707{
708 struct socket *so;
709
710 while ((so = tcb.so_next) != &tcb)
711 {
712 if (so->so_state & SS_NOFDREF || so->s == -1)
713 sofree(pData, so);
714 else
715 tcp_drop(pData, sototcpcb(so), 0);
716 }
717
718 while ((so = udb.so_next) != &udb)
719 udp_detach(pData, so);
720
721 link_up = 0;
722}
723
724/**
725 * Terminates the slirp component.
726 */
727void slirp_term(PNATState pData)
728{
729#ifndef VBOX_WITH_MULTI_DNS
730 if (pData->pszDomain)
731 RTStrFree((char *)(void *)pData->pszDomain);
732#endif
733
734#ifdef RT_OS_WINDOWS
735 pData->pfIcmpCloseHandle(pData->icmp_socket.sh);
736 FreeLibrary(pData->hmIcmpLibrary);
737 RTMemFree(pData->pvIcmpBuffer);
738# else
739 closesocket(pData->icmp_socket.s);
740#endif
741
742 slirp_link_down(pData);
743#ifdef VBOX_WITH_MULTI_DNS
744 slirp_release_dns_list(pData);
745#endif
746#ifdef RT_OS_WINDOWS
747 WSACleanup();
748#endif
749#ifdef LOG_ENABLED
750 Log(("\n"
751 "NAT statistics\n"
752 "--------------\n"
753 "\n"));
754 ipstats(pData);
755 tcpstats(pData);
756 udpstats(pData);
757 icmpstats(pData);
758 mbufstats(pData);
759 sockstats(pData);
760 Log(("\n"
761 "\n"
762 "\n"));
763#endif
764 RTMemFree(pData);
765}
766
767
768#define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
769#define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
770#define UPD_NFDS(x) do { if (nfds < (x)) nfds = (x); } while (0)
771
772/*
773 * curtime kept to an accuracy of 1ms
774 */
775static void updtime(PNATState pData)
776{
777#ifdef RT_OS_WINDOWS
778 struct _timeb tb;
779
780 _ftime(&tb);
781 curtime = (u_int)tb.time * (u_int)1000;
782 curtime += (u_int)tb.millitm;
783#else
784 gettimeofday(&tt, 0);
785
786 curtime = (u_int)tt.tv_sec * (u_int)1000;
787 curtime += (u_int)tt.tv_usec / (u_int)1000;
788
789 if ((tt.tv_usec % 1000) >= 500)
790 curtime++;
791#endif
792}
793
794#ifdef RT_OS_WINDOWS
795void slirp_select_fill(PNATState pData, int *pnfds)
796#else /* RT_OS_WINDOWS */
797void slirp_select_fill(PNATState pData, int *pnfds, struct pollfd *polls)
798#endif /* !RT_OS_WINDOWS */
799{
800 struct socket *so, *so_next;
801 int nfds;
802#if defined(RT_OS_WINDOWS)
803 int rc;
804 int error;
805#else
806 int poll_index = 0;
807#endif
808 int i;
809
810 STAM_PROFILE_START(&pData->StatFill, a);
811
812 nfds = *pnfds;
813
814 /*
815 * First, TCP sockets
816 */
817 do_slowtimo = 0;
818 if (link_up)
819 {
820 /*
821 * *_slowtimo needs calling if there are IP fragments
822 * in the fragment queue, or there are TCP connections active
823 */
824 /* XXX:
825 * triggering of fragment expiration should be the same but use new macroses
826 */
827 do_slowtimo = (tcb.so_next != &tcb);
828 if (!do_slowtimo)
829 {
830 for (i = 0; i < IPREASS_NHASH; i++)
831 {
832 if (!TAILQ_EMPTY(&ipq[i]))
833 {
834 do_slowtimo = 1;
835 break;
836 }
837 }
838 }
839 ICMP_ENGAGE_EVENT(&pData->icmp_socket, readfds);
840
841 STAM_COUNTER_RESET(&pData->StatTCP);
842 STAM_COUNTER_RESET(&pData->StatTCPHot);
843
844 QSOCKET_FOREACH(so, so_next, tcp)
845 /* { */
846#if !defined(RT_OS_WINDOWS)
847 so->so_poll_index = -1;
848#endif
849 STAM_COUNTER_INC(&pData->StatTCP);
850
851 /*
852 * See if we need a tcp_fasttimo
853 */
854 if ( time_fasttimo == 0
855 && so->so_tcpcb != NULL
856 && so->so_tcpcb->t_flags & TF_DELACK)
857 time_fasttimo = curtime; /* Flag when we want a fasttimo */
858
859 /*
860 * NOFDREF can include still connecting to local-host,
861 * newly socreated() sockets etc. Don't want to select these.
862 */
863 if (so->so_state & SS_NOFDREF || so->s == -1)
864 CONTINUE(tcp);
865
866 /*
867 * Set for reading sockets which are accepting
868 */
869 if (so->so_state & SS_FACCEPTCONN)
870 {
871 STAM_COUNTER_INC(&pData->StatTCPHot);
872 TCP_ENGAGE_EVENT1(so, readfds);
873 CONTINUE(tcp);
874 }
875
876 /*
877 * Set for writing sockets which are connecting
878 */
879 if (so->so_state & SS_ISFCONNECTING)
880 {
881 Log2(("connecting %R[natsock] engaged\n",so));
882 STAM_COUNTER_INC(&pData->StatTCPHot);
883 TCP_ENGAGE_EVENT1(so, writefds);
884 }
885
886 /*
887 * Set for writing if we are connected, can send more, and
888 * we have something to send
889 */
890 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc)
891 {
892 STAM_COUNTER_INC(&pData->StatTCPHot);
893 TCP_ENGAGE_EVENT1(so, writefds);
894 }
895
896 /*
897 * Set for reading (and urgent data) if we are connected, can
898 * receive more, and we have room for it XXX /2 ?
899 */
900 if (CONN_CANFRCV(so) && (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2)))
901 {
902 STAM_COUNTER_INC(&pData->StatTCPHot);
903 TCP_ENGAGE_EVENT2(so, readfds, xfds);
904 }
905 LOOP_LABEL(tcp, so, so_next);
906 }
907
908 /*
909 * UDP sockets
910 */
911 STAM_COUNTER_RESET(&pData->StatUDP);
912 STAM_COUNTER_RESET(&pData->StatUDPHot);
913
914 QSOCKET_FOREACH(so, so_next, udp)
915 /* { */
916
917 STAM_COUNTER_INC(&pData->StatUDP);
918#if !defined(RT_OS_WINDOWS)
919 so->so_poll_index = -1;
920#endif
921
922 /*
923 * See if it's timed out
924 */
925 if (so->so_expire)
926 {
927 if (so->so_expire <= curtime)
928 {
929#ifdef VBOX_WITH_SLIRP_DNS_PROXY
930 Log2(("NAT: %R[natsock] expired\n", so));
931 if (so->so_timeout != NULL)
932 {
933 so->so_timeout(pData, so, so->so_timeout_arg);
934 }
935#endif
936#ifdef VBOX_WITH_SLIRP_MT
937 /* we need so_next for continue our cycle*/
938 so_next = so->so_next;
939#endif
940 UDP_DETACH(pData, so, so_next);
941 CONTINUE_NO_UNLOCK(udp);
942 }
943 else
944 do_slowtimo = 1; /* Let socket expire */
945 }
946
947 /*
948 * When UDP packets are received from over the link, they're
949 * sendto()'d straight away, so no need for setting for writing
950 * Limit the number of packets queued by this session to 4.
951 * Note that even though we try and limit this to 4 packets,
952 * the session could have more queued if the packets needed
953 * to be fragmented.
954 *
955 * (XXX <= 4 ?)
956 */
957 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4)
958 {
959 STAM_COUNTER_INC(&pData->StatUDPHot);
960 UDP_ENGAGE_EVENT(so, readfds);
961 }
962 LOOP_LABEL(udp, so, so_next);
963 }
964
965 }
966
967#if defined(RT_OS_WINDOWS)
968 *pnfds = VBOX_EVENT_COUNT;
969#else /* RT_OS_WINDOWS */
970 AssertRelease(poll_index <= *pnfds);
971 *pnfds = poll_index;
972#endif /* !RT_OS_WINDOWS */
973
974 STAM_PROFILE_STOP(&pData->StatFill, a);
975}
976
977#if defined(RT_OS_WINDOWS)
978void slirp_select_poll(PNATState pData, int fTimeout, int fIcmp)
979#else /* RT_OS_WINDOWS */
980void slirp_select_poll(PNATState pData, struct pollfd *polls, int ndfs)
981#endif /* !RT_OS_WINDOWS */
982{
983 struct socket *so, *so_next;
984 int ret;
985#if defined(RT_OS_WINDOWS)
986 WSANETWORKEVENTS NetworkEvents;
987 int rc;
988 int error;
989#else
990 int poll_index = 0;
991#endif
992
993 STAM_PROFILE_START(&pData->StatPoll, a);
994
995 /* Update time */
996 updtime(pData);
997
998 /*
999 * See if anything has timed out
1000 */
1001 if (link_up)
1002 {
1003 if (time_fasttimo && ((curtime - time_fasttimo) >= 2))
1004 {
1005 STAM_PROFILE_START(&pData->StatFastTimer, a);
1006 tcp_fasttimo(pData);
1007 time_fasttimo = 0;
1008 STAM_PROFILE_STOP(&pData->StatFastTimer, a);
1009 }
1010 if (do_slowtimo && ((curtime - last_slowtimo) >= 499))
1011 {
1012 STAM_PROFILE_START(&pData->StatSlowTimer, a);
1013 ip_slowtimo(pData);
1014 tcp_slowtimo(pData);
1015 last_slowtimo = curtime;
1016 STAM_PROFILE_STOP(&pData->StatSlowTimer, a);
1017 }
1018 }
1019#if defined(RT_OS_WINDOWS)
1020 if (fTimeout)
1021 return; /* only timer update */
1022#endif
1023
1024 /*
1025 * Check sockets
1026 */
1027 if (link_up)
1028 {
1029#if defined(RT_OS_WINDOWS)
1030 /*XXX: before renaming please make see define
1031 * fIcmp in slirp_state.h
1032 */
1033 if (fIcmp)
1034 sorecvfrom(pData, &pData->icmp_socket);
1035#else
1036 if ( (pData->icmp_socket.s != -1)
1037 && CHECK_FD_SET(&pData->icmp_socket, ignored, readfds))
1038 sorecvfrom(pData, &pData->icmp_socket);
1039#endif
1040 /*
1041 * Check TCP sockets
1042 */
1043 QSOCKET_FOREACH(so, so_next, tcp)
1044 /* { */
1045
1046#ifdef VBOX_WITH_SLIRP_MT
1047 if ( so->so_state & SS_NOFDREF
1048 && so->so_deleted == 1)
1049 {
1050 struct socket *son, *sop = NULL;
1051 QSOCKET_LOCK(tcb);
1052 if (so->so_next != NULL)
1053 {
1054 if (so->so_next != &tcb)
1055 SOCKET_LOCK(so->so_next);
1056 son = so->so_next;
1057 }
1058 if ( so->so_prev != &tcb
1059 && so->so_prev != NULL)
1060 {
1061 SOCKET_LOCK(so->so_prev);
1062 sop = so->so_prev;
1063 }
1064 QSOCKET_UNLOCK(tcb);
1065 remque(pData, so);
1066 NSOCK_DEC();
1067 SOCKET_UNLOCK(so);
1068 SOCKET_LOCK_DESTROY(so);
1069 RTMemFree(so);
1070 so_next = son;
1071 if (sop != NULL)
1072 SOCKET_UNLOCK(sop);
1073 CONTINUE_NO_UNLOCK(tcp);
1074 }
1075#endif
1076 /*
1077 * FD_ISSET is meaningless on these sockets
1078 * (and they can crash the program)
1079 */
1080 if (so->so_state & SS_NOFDREF || so->s == -1)
1081 CONTINUE(tcp);
1082
1083 POLL_TCP_EVENTS(rc, error, so, &NetworkEvents);
1084
1085 LOG_NAT_SOCK(so, TCP, &NetworkEvents, readfds, writefds, xfds);
1086
1087
1088 /*
1089 * Check for URG data
1090 * This will soread as well, so no need to
1091 * test for readfds below if this succeeds
1092 */
1093
1094 /* out-of-band data */
1095 if (CHECK_FD_SET(so, NetworkEvents, xfds))
1096 {
1097 sorecvoob(pData, so);
1098 }
1099
1100 /*
1101 * Check sockets for reading
1102 */
1103 else if ( CHECK_FD_SET(so, NetworkEvents, readfds)
1104 || WIN_CHECK_FD_SET(so, NetworkEvents, acceptds))
1105 {
1106 /*
1107 * Check for incoming connections
1108 */
1109 if (so->so_state & SS_FACCEPTCONN)
1110 {
1111 TCP_CONNECT(pData, so);
1112#if defined(RT_OS_WINDOWS)
1113 if (!(NetworkEvents.lNetworkEvents & FD_CLOSE))
1114#endif
1115 CONTINUE(tcp);
1116 }
1117
1118 ret = soread(pData, so);
1119 /* Output it if we read something */
1120 if (ret > 0)
1121 TCP_OUTPUT(pData, sototcpcb(so));
1122 }
1123
1124#if defined(RT_OS_WINDOWS)
1125 /*
1126 * Check for FD_CLOSE events.
1127 * in some cases once FD_CLOSE engaged on socket it could be flashed latter (for some reasons)
1128 */
1129 if ( (NetworkEvents.lNetworkEvents & FD_CLOSE)
1130 || (so->so_close == 1))
1131 {
1132 so->so_close = 1; /* mark it */
1133 /*
1134 * drain the socket
1135 */
1136 for (;;)
1137 {
1138 ret = soread(pData, so);
1139 if (ret > 0)
1140 TCP_OUTPUT(pData, sototcpcb(so));
1141 else
1142 break;
1143 }
1144 CONTINUE(tcp);
1145 }
1146#endif
1147
1148 /*
1149 * Check sockets for writing
1150 */
1151 if (CHECK_FD_SET(so, NetworkEvents, writefds))
1152 {
1153 /*
1154 * Check for non-blocking, still-connecting sockets
1155 */
1156 if (so->so_state & SS_ISFCONNECTING)
1157 {
1158 Log2(("connecting %R[natsock] catched\n", so));
1159 /* Connected */
1160 so->so_state &= ~SS_ISFCONNECTING;
1161
1162 /*
1163 * This should be probably guarded by PROBE_CONN too. Anyway,
1164 * we disable it on OS/2 because the below send call returns
1165 * EFAULT which causes the opened TCP socket to close right
1166 * after it has been opened and connected.
1167 */
1168#ifndef RT_OS_OS2
1169 ret = send(so->s, (const char *)&ret, 0, 0);
1170 if (ret < 0)
1171 {
1172 /* XXXXX Must fix, zero bytes is a NOP */
1173 if ( errno == EAGAIN
1174 || errno == EWOULDBLOCK
1175 || errno == EINPROGRESS
1176 || errno == ENOTCONN)
1177 CONTINUE(tcp);
1178
1179 /* else failed */
1180 so->so_state = SS_NOFDREF;
1181 }
1182 /* else so->so_state &= ~SS_ISFCONNECTING; */
1183#endif
1184
1185 /*
1186 * Continue tcp_input
1187 */
1188 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1189 /* continue; */
1190 }
1191 else
1192 SOWRITE(ret, pData, so);
1193 /*
1194 * XXX If we wrote something (a lot), there could be the need
1195 * for a window update. In the worst case, the remote will send
1196 * a window probe to get things going again.
1197 */
1198 }
1199
1200 /*
1201 * Probe a still-connecting, non-blocking socket
1202 * to check if it's still alive
1203 */
1204#ifdef PROBE_CONN
1205 if (so->so_state & SS_ISFCONNECTING)
1206 {
1207 ret = recv(so->s, (char *)&ret, 0, 0);
1208
1209 if (ret < 0)
1210 {
1211 /* XXX */
1212 if ( errno == EAGAIN
1213 || errno == EWOULDBLOCK
1214 || errno == EINPROGRESS
1215 || errno == ENOTCONN)
1216 {
1217 CONTINUE(tcp); /* Still connecting, continue */
1218 }
1219
1220 /* else failed */
1221 so->so_state = SS_NOFDREF;
1222
1223 /* tcp_input will take care of it */
1224 }
1225 else
1226 {
1227 ret = send(so->s, &ret, 0, 0);
1228 if (ret < 0)
1229 {
1230 /* XXX */
1231 if ( errno == EAGAIN
1232 || errno == EWOULDBLOCK
1233 || errno == EINPROGRESS
1234 || errno == ENOTCONN)
1235 {
1236 CONTINUE(tcp);
1237 }
1238 /* else failed */
1239 so->so_state = SS_NOFDREF;
1240 }
1241 else
1242 so->so_state &= ~SS_ISFCONNECTING;
1243
1244 }
1245 TCP_INPUT((struct mbuf *)NULL, sizeof(struct ip),so);
1246 } /* SS_ISFCONNECTING */
1247#endif
1248#ifndef RT_OS_WINDOWS
1249 if ( UNIX_CHECK_FD_SET(so, NetworkEvents, rdhup)
1250 || UNIX_CHECK_FD_SET(so, NetworkEvents, rderr))
1251 {
1252 int err;
1253 int inq, outq;
1254 int status;
1255 socklen_t optlen = sizeof(int);
1256 inq = outq = 0;
1257 status = getsockopt(so->s, SOL_SOCKET, SO_ERROR, &err, &optlen);
1258 if (status != 0)
1259 Log(("NAT: can't get error status from %R[natsock]\n", so));
1260#ifndef RT_OS_SOLARIS
1261 status = ioctl(so->s, FIONREAD, &inq); /* tcp(7) recommends SIOCINQ which is Linux specific */
1262 if (status != 0 || status != EINVAL)
1263 {
1264 /* EINVAL returned if socket in listen state tcp(7)*/
1265 Log(("NAT: can't get depth of IN queue status from %R[natsock]\n", so));
1266 }
1267 status = ioctl(so->s, TIOCOUTQ, &outq); /* SIOCOUTQ see previous comment */
1268 if (status != 0)
1269 Log(("NAT: can't get depth of OUT queue from %R[natsock]\n", so));
1270#else
1271 /*
1272 * Solaris has bit different ioctl commands and its handlings
1273 * hint: streamio(7) I_NREAD
1274 */
1275#endif
1276 if ( so->so_state & SS_ISFCONNECTING
1277 || UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1278 {
1279 /**
1280 * Check if we need here take care about gracefull connection
1281 * @todo try with proxy server
1282 */
1283 if (UNIX_CHECK_FD_SET(so, NetworkEvents, readfds))
1284 {
1285 /*
1286 * Never meet inq != 0 or outq != 0, anyway let it stay for a while
1287 * in case it happens we'll able to detect it.
1288 * Give TCP/IP stack wait or expire the socket.
1289 */
1290 Log(("NAT: %R[natsock] err(%d:%s) s(in:%d,out:%d)happens on read I/O, "
1291 "other side close connection \n", so, err, strerror(err), inq, outq));
1292 CONTINUE(tcp);
1293 }
1294 goto tcp_input_close;
1295 }
1296 if ( !UNIX_CHECK_FD_SET(so, NetworkEvents, readfds)
1297 && !UNIX_CHECK_FD_SET(so, NetworkEvents, writefds)
1298 && !UNIX_CHECK_FD_SET(so, NetworkEvents, xfds))
1299 {
1300 Log(("NAT: system expires the socket %R[natsock] err(%d:%s) s(in:%d,out:%d) happens on non-I/O. ",
1301 so, err, strerror(err), inq, outq));
1302 goto tcp_input_close;
1303 }
1304 Log(("NAT: %R[natsock] we've met(%d:%s) s(in:%d, out:%d) unhandled combination hup (%d) "
1305 "rederr(%d) on (r:%d, w:%d, x:%d)\n",
1306 so, err, strerror(err),
1307 inq, outq,
1308 UNIX_CHECK_FD_SET(so, ign, rdhup),
1309 UNIX_CHECK_FD_SET(so, ign, rderr),
1310 UNIX_CHECK_FD_SET(so, ign, readfds),
1311 UNIX_CHECK_FD_SET(so, ign, writefds),
1312 UNIX_CHECK_FD_SET(so, ign, xfds)));
1313 /*
1314 * Give OS's TCP/IP stack a chance to resolve an issue or expire the socket.
1315 */
1316 CONTINUE(tcp);
1317tcp_input_close:
1318 so->so_state = SS_NOFDREF; /*cause connection valid tcp connection termination and socket closing */
1319 TCP_INPUT(pData, (struct mbuf *)NULL, sizeof(struct ip), so);
1320 CONTINUE(tcp);
1321 }
1322#endif
1323 LOOP_LABEL(tcp, so, so_next);
1324 }
1325
1326 /*
1327 * Now UDP sockets.
1328 * Incoming packets are sent straight away, they're not buffered.
1329 * Incoming UDP data isn't buffered either.
1330 */
1331 QSOCKET_FOREACH(so, so_next, udp)
1332 /* { */
1333#ifdef VBOX_WITH_SLIRP_MT
1334 if ( so->so_state & SS_NOFDREF
1335 && so->so_deleted == 1)
1336 {
1337 struct socket *son, *sop = NULL;
1338 QSOCKET_LOCK(udb);
1339 if (so->so_next != NULL)
1340 {
1341 if (so->so_next != &udb)
1342 SOCKET_LOCK(so->so_next);
1343 son = so->so_next;
1344 }
1345 if ( so->so_prev != &udb
1346 && so->so_prev != NULL)
1347 {
1348 SOCKET_LOCK(so->so_prev);
1349 sop = so->so_prev;
1350 }
1351 QSOCKET_UNLOCK(udb);
1352 remque(pData, so);
1353 NSOCK_DEC();
1354 SOCKET_UNLOCK(so);
1355 SOCKET_LOCK_DESTROY(so);
1356 RTMemFree(so);
1357 so_next = son;
1358 if (sop != NULL)
1359 SOCKET_UNLOCK(sop);
1360 CONTINUE_NO_UNLOCK(udp);
1361 }
1362#endif
1363 POLL_UDP_EVENTS(rc, error, so, &NetworkEvents);
1364
1365 LOG_NAT_SOCK(so, UDP, &NetworkEvents, readfds, writefds, xfds);
1366
1367 if (so->s != -1 && CHECK_FD_SET(so, NetworkEvents, readfds))
1368 {
1369 SORECVFROM(pData, so);
1370 }
1371 LOOP_LABEL(udp, so, so_next);
1372 }
1373
1374 }
1375
1376#ifndef VBOX_WITH_SLIRP_MT
1377 /*
1378 * See if we can start outputting
1379 */
1380 if (if_queued && link_up)
1381 if_start(pData);
1382#endif
1383
1384 STAM_PROFILE_STOP(&pData->StatPoll, a);
1385}
1386
1387#ifndef VBOX_WITH_NAT_SERVICE
1388#define ETH_ALEN 6
1389#define ETH_HLEN 14
1390
1391#define ARPOP_REQUEST 1 /* ARP request */
1392#define ARPOP_REPLY 2 /* ARP reply */
1393
1394struct ethhdr
1395{
1396 unsigned char h_dest[ETH_ALEN]; /* destination eth addr */
1397 unsigned char h_source[ETH_ALEN]; /* source ether addr */
1398 unsigned short h_proto; /* packet type ID field */
1399};
1400AssertCompileSize(struct ethhdr, 14);
1401#endif
1402
1403struct arphdr
1404{
1405 unsigned short ar_hrd; /* format of hardware address */
1406 unsigned short ar_pro; /* format of protocol address */
1407 unsigned char ar_hln; /* length of hardware address */
1408 unsigned char ar_pln; /* length of protocol address */
1409 unsigned short ar_op; /* ARP opcode (command) */
1410
1411 /*
1412 * Ethernet looks like this : This bit is variable sized however...
1413 */
1414 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
1415 unsigned char ar_sip[4]; /* sender IP address */
1416 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
1417 unsigned char ar_tip[4]; /* target IP address */
1418};
1419AssertCompileSize(struct arphdr, 28);
1420
1421static void arp_input(PNATState pData, struct mbuf *m)
1422{
1423 struct ethhdr *eh;
1424 struct ethhdr *reh;
1425 struct arphdr *ah;
1426 struct arphdr *rah;
1427 int ar_op;
1428 struct ex_list *ex_ptr;
1429 uint32_t htip;
1430 uint32_t tip;
1431 struct mbuf *mr;
1432 eh = mtod(m, struct ethhdr *);
1433 ah = (struct arphdr *)&eh[1];
1434 htip = ntohl(*(uint32_t*)ah->ar_tip);
1435 tip = *(uint32_t*)ah->ar_tip;
1436
1437 mr = m_get(pData);
1438#ifdef VBOX_WITH_NAT_SERVICE
1439 reh = mtod(mr, struct ethhdr *);
1440 memcpy(reh->h_source, eh->h_source, ETH_ALEN); /* XXX: if_encap will swap src and dst*/
1441 Log4(("NAT: arp:[%hhx:%hhx:%hhx:%hhx:%hhx:%hhx]->[%hhx:%hhx:%hhx:%hhx:%hhx:%hhx]\n",
1442 reh->h_source[0], reh->h_source[1], reh->h_source[2], reh->h_source[3], reh->h_source[4], reh->h_source[5],
1443 reh->h_dest[0], reh->h_dest[1], reh->h_dest[2], reh->h_dest[3], reh->h_dest[4], reh->h_dest[5]));
1444 Log4(("NAT: arp: %R[IP4]\n", &tip));
1445#endif
1446 mr->m_data += if_maxlinkhdr;
1447 mr->m_len = sizeof(struct arphdr);
1448 rah = mtod(mr, struct arphdr *);
1449
1450 ar_op = ntohs(ah->ar_op);
1451 switch(ar_op)
1452 {
1453 case ARPOP_REQUEST:
1454#ifdef VBOX_WITH_NAT_SERVICE
1455 if (tip == special_addr.s_addr) goto arp_ok;
1456#endif
1457 if ((htip & pData->netmask) == ntohl(special_addr.s_addr))
1458 {
1459 if ( CTL_CHECK(htip, CTL_DNS)
1460 || CTL_CHECK(htip, CTL_ALIAS)
1461 || CTL_CHECK(htip, CTL_TFTP))
1462 goto arp_ok;
1463 for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1464 {
1465 if ((htip & ~pData->netmask) == ex_ptr->ex_addr)
1466 {
1467 goto arp_ok;
1468 }
1469 }
1470 return;
1471 arp_ok:
1472 rah->ar_hrd = htons(1);
1473 rah->ar_pro = htons(ETH_P_IP);
1474 rah->ar_hln = ETH_ALEN;
1475 rah->ar_pln = 4;
1476 rah->ar_op = htons(ARPOP_REPLY);
1477 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN);
1478
1479 switch (htip & ~pData->netmask)
1480 {
1481 case CTL_DNS:
1482 case CTL_ALIAS:
1483 rah->ar_sha[5] = (uint8_t)(htip & ~pData->netmask);
1484 break;
1485 default:;
1486 }
1487
1488 memcpy(rah->ar_sip, ah->ar_tip, 4);
1489 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
1490 memcpy(rah->ar_tip, ah->ar_sip, 4);
1491 if_encap(pData, ETH_P_ARP, mr);
1492 m_free(pData, m);
1493 }
1494 break;
1495 default:
1496 break;
1497 }
1498}
1499
1500void slirp_input(PNATState pData, const uint8_t *pkt, int pkt_len)
1501{
1502 struct mbuf *m;
1503 int proto;
1504 static bool fWarnedIpv6;
1505 struct ethhdr *eh = (struct ethhdr*)pkt;
1506
1507 if (pkt_len < ETH_HLEN)
1508 {
1509 LogRel(("NAT: packet having size %d has been ingnored\n", pkt_len));
1510 return;
1511 }
1512 Log4(("NAT: in:[%hhx:%hhx:%hhx:%hhx:%hhx:%hhx]->[%hhx:%hhx:%hhx:%hhx:%hhx:%hhx]\n",
1513 eh->h_source[0], eh->h_source[1], eh->h_source[2], eh->h_source[3], eh->h_source[4], eh->h_source[5],
1514 eh->h_dest[0], eh->h_dest[1], eh->h_dest[2], eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]));
1515#ifdef VBOX_WITH_NAT_SERVICE
1516 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) == 0)
1517 {
1518 goto drop;
1519 }
1520#endif
1521
1522 m = m_get(pData);
1523 if (!m)
1524 {
1525 LogRel(("NAT: can't allocate new mbuf\n"));
1526 return;
1527 }
1528
1529 /* Note: we add to align the IP header */
1530
1531 if (M_FREEROOM(m) < pkt_len)
1532 m_inc(m, pkt_len);
1533
1534 m->m_len = pkt_len ;
1535 memcpy(m->m_data, pkt, pkt_len);
1536
1537 proto = ntohs(*(uint16_t *)(pkt + 12));
1538 switch(proto)
1539 {
1540 case ETH_P_ARP:
1541 arp_input(pData, m);
1542 break;
1543 case ETH_P_IP:
1544 /* Update time. Important if the network is very quiet, as otherwise
1545 * the first outgoing connection gets an incorrect timestamp. */
1546 updtime(pData);
1547 m->m_data += ETH_HLEN;
1548 m->m_len -= ETH_HLEN;
1549 ip_input(pData, m);
1550 break;
1551 case ETH_P_IPV6:
1552 m_free(pData, m);
1553 if (!fWarnedIpv6)
1554 {
1555 LogRel(("NAT: IPv6 not supported\n"));
1556 fWarnedIpv6 = true;
1557 }
1558 break;
1559 default:
1560 LogRel(("NAT: Unsupported protocol %x\n", proto));
1561 m_free(pData, m);
1562 break;
1563 }
1564 drop:
1565 RTMemFree((void *)pkt);
1566}
1567
1568/* output the IP packet to the ethernet device */
1569void if_encap(PNATState pData, uint16_t eth_proto, struct mbuf *m)
1570{
1571 struct ethhdr *eh;
1572 uint8_t *buf = RTMemAlloc(1600);
1573
1574 m->m_data -= if_maxlinkhdr;
1575 m->m_len += ETH_HLEN;
1576 eh = mtod(m, struct ethhdr *);
1577
1578#ifndef VBOX_WITH_NAT_SERVICE
1579 memcpy(eh->h_dest, client_ethaddr, ETH_ALEN);
1580 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 1);
1581 /* XXX: not correct */
1582 eh->h_source[5] = CTL_ALIAS;
1583#else
1584 Assert((caddr_t)eh == (caddr_t)m->m_dat);
1585 if (memcmp(eh->h_source, special_ethaddr, ETH_ALEN) != 0)
1586 {
1587 memcpy(eh->h_dest, eh->h_source, ETH_ALEN);
1588 memcpy(eh->h_source, special_ethaddr, ETH_ALEN);
1589 Assert(memcmp(eh->h_dest, special_ethaddr, ETH_ALEN) != 0);
1590 Assert(memcmp(eh->h_dest, zerro_ethaddr, ETH_ALEN) != 0);
1591 }
1592#endif
1593 eh->h_proto = htons(eth_proto);
1594#if 0
1595 slirp_output(pData->pvUser, m, mtod(m, uint8_t *), m->m_len);
1596#else
1597 memcpy(buf, mtod(m, uint8_t *), m->m_len);
1598 slirp_output(pData->pvUser, NULL, buf, m->m_len);
1599 m_free(pData, m);
1600#endif
1601}
1602
1603int slirp_redir(PNATState pData, int is_udp, int host_port,
1604 struct in_addr guest_addr, int guest_port)
1605{
1606 struct socket *so;
1607 Log2(("NAT: set redirect %s hp:%d gp:%d\n", (is_udp?"UDP":"TCP"), host_port, guest_port));
1608 if (is_udp)
1609 {
1610 so = udp_listen(pData, htons(host_port), guest_addr.s_addr,
1611 htons(guest_port), 0);
1612 }
1613 else
1614 {
1615 so = solisten(pData, htons(host_port), guest_addr.s_addr,
1616 htons(guest_port), 0);
1617 }
1618 Log2(("NAT: redirecting socket %R[natsock]\n", so));
1619 return (so != NULL ? 0 : -1);
1620}
1621
1622int slirp_add_exec(PNATState pData, int do_pty, const char *args, int addr_low_byte,
1623 int guest_port)
1624{
1625 return add_exec(&exec_list, do_pty, (char *)args,
1626 addr_low_byte, htons(guest_port));
1627}
1628
1629void slirp_set_ethaddr(PNATState pData, const uint8_t *ethaddr)
1630{
1631#ifndef VBOX_WITH_NAT_SERVICE
1632 memcpy(client_ethaddr, ethaddr, ETH_ALEN);
1633#endif
1634}
1635
1636#if defined(RT_OS_WINDOWS)
1637HANDLE *slirp_get_events(PNATState pData)
1638{
1639 return pData->phEvents;
1640}
1641void slirp_register_external_event(PNATState pData, HANDLE hEvent, int index)
1642{
1643 pData->phEvents[index] = hEvent;
1644}
1645#endif
1646
1647unsigned int slirp_get_timeout_ms(PNATState pData)
1648{
1649 if (link_up)
1650 {
1651 if (time_fasttimo)
1652 return 2;
1653 if (do_slowtimo)
1654 return 500; /* see PR_SLOWHZ */
1655 }
1656 return 0;
1657}
1658
1659#ifndef RT_OS_WINDOWS
1660int slirp_get_nsock(PNATState pData)
1661{
1662 return pData->nsock;
1663}
1664#endif
1665
1666/*
1667 * this function called from NAT thread
1668 */
1669void slirp_post_sent(PNATState pData, void *pvArg)
1670{
1671 struct socket *so = 0;
1672 struct tcpcb *tp = 0;
1673 struct mbuf *m = (struct mbuf *)pvArg;
1674 m_free(pData, m);
1675}
1676#ifdef VBOX_WITH_SLIRP_MT
1677void slirp_process_queue(PNATState pData)
1678{
1679 RTReqProcess(pData->pReqQueue, RT_INDEFINITE_WAIT);
1680}
1681void *slirp_get_queue(PNATState pData)
1682{
1683 return pData->pReqQueue;
1684}
1685#endif
1686
1687uint16_t slirp_get_service(int proto, uint16_t dport, uint16_t sport)
1688{
1689 uint16_t hdport, hsport, service;
1690 hdport = ntohs(dport);
1691 hsport = ntohs(sport);
1692 Log2(("proto: %d, dport: %d sport: %d\n", proto, hdport, hsport));
1693 service = 0;
1694#if 0
1695 /* Always return 0 here */
1696 switch (hdport)
1697 {
1698 case 500:
1699 /* service = sport; */
1700 break;
1701 }
1702#endif
1703 Log2(("service : %d\n", service));
1704 return htons(service);
1705}
1706
1707void slirp_set_dhcp_TFTP_prefix(PNATState pData, const char *tftpPrefix)
1708{
1709 Log2(("tftp_prefix:%s\n", tftpPrefix));
1710 tftp_prefix = tftpPrefix;
1711}
1712
1713void slirp_set_dhcp_TFTP_bootfile(PNATState pData, const char *bootFile)
1714{
1715 Log2(("bootFile:%s\n", bootFile));
1716 bootp_filename = bootFile;
1717}
1718
1719void slirp_set_dhcp_next_server(PNATState pData, const char *next_server)
1720{
1721 Log2(("next_server:%s\n", next_server));
1722 if (next_server == NULL)
1723 pData->tftp_server.s_addr = htonl(ntohl(special_addr.s_addr) | CTL_TFTP);
1724 else
1725 inet_aton(next_server, &pData->tftp_server);
1726}
1727#ifdef VBOX_WITH_SLIRP_DNS_PROXY
1728void slirp_set_dhcp_dns_proxy(PNATState pData, bool fDNSProxy)
1729{
1730 Log2(("NAT: DNS proxy switched %s\n", (fDNSProxy ? "on" : "off")));
1731 pData->use_dns_proxy = fDNSProxy;
1732}
1733#endif
Note: See TracBrowser for help on using the repository browser.

© 2025 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette